main.py

from __future__ import with_statement
import json
import os
from os.path import basename, splitext

from sklearn.externals import joblib

from clustering import cluster
from posture_classifier import Example, tree_classifier, filter_by_kl
from skeleton_features import all_features

valid_sensors = [] #empty means all sensors
raw_data_path = '/home/ami/cosmin/proiecte/posture-recognition/raw-data-test'


def example_from_file(filename, label = None):
    '''Returns an Experiment object reading example skeletons from an input file
       if label is None than the file name (w/o extension) is used
    '''
    
    print 'Processing file', filename
    if label is None:
        label = splitext(basename(filename))[0]
    
    with open(filename) as f:
        objects = [json.loads(line) for line in f]
        examples = [all_features(obj['skeleton_3D']) for obj in objects if validJson(obj)]
        return Example(label, examples)

def main():
    def validFile(filename):
        return splitext(filename)[1] == '.txt'    

    if not os.path.isdir(raw_data_path):
        raise "%s is not a folder" % raw_data_path
    
    examples = [example_from_file(raw_data_path + '/' + filename) for filename in os.listdir(raw_data_path) if validFile(filename)]
    
#     classifier = tree_classifier(examples)    
#     filename = '/tmp/digits_classifier.joblib.pkl'
#     _ = joblib.dump(classifier, filename, compress=9)
#         
#     print classifier

    for e in examples:
        print cluster(e.feature_vectors)
        
    
def validJson(m):
    '''Returns true iff the json is of interest'''
    if (m['type'] != 'skeleton'):
        return False;
    elif len(valid_sensors) > 0 and  (m['sensor_id'] not in valid_sensors):
        return False
    else:
        return True



if __name__ == "__main__":
    main()